[0:00:00] For the other people, they'll join us online. But we probably all know you better than you know who we are. Just as a little reminder, we are a venture builder, a studio. We created nine companies up to now, SaaS-based companies. And you probably know two of them a little bit better than the others. They went to YC, Mathilde and...
[0:00:26] Pauline, who we recently met for a couple of months. So, short speech, because it's important to leave the floor to you. That's why we're all here. So, thank you for coming, and guys, enjoy it. So, I apologize, I've lost my voice. I will try to speak as loudly as I can. I thought we'd talk a little bit about Y Combinator and how we work and what we've learned about startups. And then, mostly, I want to just take questions. And I'm happy...
[0:01:02] In questions, talk about anything I work on. So in addition to Y Combinator, I help run OpenAI. We're trying to develop superhuman artificial intelligence and have it go as well as it possibly can for humanity. I work on basic income and other projects related to what happens when humans have a lot of their work replaced by machines. And then I work, I'm very interested in clean energy, especially nuclear.
[0:01:31] And synthetic biology. So happy to talk about any of those things. Y Combinator's been around for 12 years. We funded 1,500 companies. They're worth about $90 billion in total. And these companies range from software companies that we're sort of best known for, like Dropbox, Airbnb, and Reddit, and Stripe, to supersonic airplane companies, like Boom, that we funded last year, and Ginkgo Bioworks, which is using synthetic biology to manufacture chemicals.
[0:02:00] Many other categories. We have nearly 20,000 companies a year apply to our programs. About 15,000 of those to our main program. And they come from all over the world. Half the companies we fund are from the US. Half the companies are from everywhere else. And over time, I'm sure the US will be a small minority. We work with companies for 10 weeks. 10 months will be a long time. And they're always amazed how much progress they can make in a short period of time.
[0:02:29] So one of the things about Y Combinator that works surprisingly well is if you take really talented people with a very clear, bold vision for the future and tell them in 10 weeks you have to stand up in front of thousands of people and tell them how you've done and it's going to be really embarrassing if you're not doing very well, people accomplish amazing things. So we've built this community of founders that help each other.
[0:02:58] Partners, there's 16 of us that advise the companies. And one of the things that we've learned is that a lot of the experience of any startup is transferable to any other startup. And so we have built this platform that is able to help startups working in lots of different areas. And in fact, the bigger we get, the better our community and our network gets. At this point, you can find a YC founder to help you with almost anything you need. We have our own internal sort of
[0:03:26] We really like companies that have ambitious visions. One of the things that has been very counterintuitive to me, but very true, is that it is easier to start a hard startup than an easy one.
[0:03:55] 12,000 photo sharing company. No one cares. It's difficult to get people excited for hiring, for investors, for press. You just won't seem that differentiated. If you're doing something that's really different and really important, then people care a lot. So, you know, a nuclear fusion company that I'm involved with can hire any software engineer they want. OpenAI, we can hire almost any software engineer we want. So I think one of the secrets is that startups
[0:04:24] At least in the US, and this is a little bit less true elsewhere in the world, but at least in the US, the more ambitious you are, the easier it is. And this is, I think, a big secret and deeply counterintuitive, and still something that most people get wrong. This is, by the way, I think, one of the things that is magical about Silicon Valley. In fact, I think there are only two things that are special about Silicon Valley. One is that we have a higher density of investors than anywhere in the world, and so companies can raise capital easily, or more easily. And the other is that...
[0:04:54] Silicon Valley is still a place, the only place I really know of, where people that have strong and somewhat crazy beliefs about the future are still taken seriously. I think this used to be true in a lot more of the world and it's now harder to find. And, you know, I remember, well, let me caveat this by saying that Paris is my favorite city in the world. If I had to move from San Francisco, I would move here. So I come here a lot. And the last time I was here,
[0:05:25] I was out to dinner with a friend, and he brought two other friends that I didn't know, and OpenAI had just launched, or a few months earlier had launched. And this was before we really had any traction. We had a small number of people. We hadn't produced any results yet. And someone said to me, one of these people was like, so you woke up one morning, and you decided that you were going to start a company to solve artificial intelligence and save the world.
[0:05:53] And I was drinking a glass of wine and I could tell this wasn't going to go well, but I didn't know how to respond. And I was like, well, it sounds bad when you say it that way, but yeah, I guess so. And he was like, you Americans are all so arrogant. I just can't deal with you. And there's some truth to that, but there's also something that sometimes leads to great results. So the downside is that people are arrogant in Silicon Valley. The upside, though, is that sometimes people accomplish really...
[0:06:25] Breathtaking things. And if you have a culture that supports that, if when you say I'm going to build a rocket company or a nuclear fusion company or a supersonic airplane company or I'm going to build the best search engine in the world, even though there's already 12 of them, if people make fun of you to your face or behind your back, you never get the outsized results. And if people have seen it work enough times,
[0:06:51] That they've learned it's a very expensive mistake to write that off and make fun of YouTube quickly. Amazing things happen sometimes, and a lot of times people fail. But so, Silicon Valley's ability to believe the unbelievable, or at least suspend disbelief on the unbelievable, I think is our other strength. So that's the only thing that I think differentiates Silicon Valley from anywhere else. Those are the only two things. My expectation is that we will see
[0:07:22] Lots and lots more startups around the world, which is why I do these trips, and it's always really fun. And we're very thankful to have funded two eFounders companies, and hope to fund a lot more. But the thing that we started with 12 years ago, and the thing that I'm convinced now is more true than ever before, is that startups are the best way to make innovation happen. And I think this gets more true every year. As the tool chain gets better, you can do more and more with less people and less money.
[0:07:52] And I think we will see things that surprise us over time. One thing I noticed when I studied the history of startups is that they sort of go in waves. So when the Internet came in 98, 99, there was then a big wave of all the Internet companies. And then when mobile came in 2007, 2008, after that, in the next few years, we had...
[0:08:19] Uber and Airbnb and Snapchat and WhatsApp and Instagram and many other companies. But basically what happens is a new platform comes that's going to grow really fast. And when that happens, but before everybody else realizes it, is this magic time to start startups. And there's one in 99. There's one in, let's say, 2008. And I think we're now just entering another one, which is going to be the machine learning wave. And I think this is sort of the next technical platform. I think the easiest layup right now in all of startups.
[0:08:50] Just to start a machine learning company for any vertical. Self-driving cars, healthcare, city planning, whatever you want. Lawyers. And you can build a lot of big companies. So I'm really excited because these periods when the great companies get started are the most fun time to be an investor and I think a founder. So in the interest of limited time, I will shut up and answer questions. And I'm happy to talk about anything anyone's interested in. Or if no one has, I can keep talking.
[0:09:22] Yeah, I have a question. Sure. You say one of the current field of study today is basic income. Yeah. Now we have elections, there are presidential elections coming in the next month in France, and it's also a theme we could hear about. But what is surprising is that the politicians that talk about basic income in France are more left-wing, socialist people now, and the people from right, like capitalists, they don't talk about basic income.
[0:09:50] Well, first of all, the political landscape in the US is
[0:10:23] Very different than the political landscape here. So the people that are sort of the capitalists are also usually the most on the left, the most liberal, the most socially liberal, often in favor of the highest taxes. Basic income, though, has managed to appeal across the entire US political spectrum in a way that's very unusual. So you have our version of socialism, whatever you want to call that, and then you have even our real conservatives who just want a smaller government.
[0:10:54] And would rather the simplicity of just give money and have the government not be involved in how it gets spent. It's been one of these very rare ideas that has united the political spectrum. I think it is still something that has the most fervent support in the tech community because we can all sense what's happening, right? So in the next 10, 20, maybe 30 years, I think almost all repetitive work done by humans where the interpersonal interaction isn't the most important part will be done by machines.
[0:11:24] And we've had these revolutions before. We had the industrial revolution, the agricultural revolution. This will be the third on that order where most human jobs change in a very short period of time. These are big. These reshape society, economies, politics, the way we interact with each other. And I think people in tech can now sense that one is coming. We don't even need any more great breakthroughs in artificial intelligence to have a massive wave of automation.
[0:11:53] And I'm sure we'll get through it. I think human demand is limitless. We'll find jobs on the other side. We'll find things to do with our time. But it's a big change. And I think people in the tech industry realize that this huge change is coming and we'll need a new social structure to deal with it. I don't think basic income appeals to everyone. And even I have questions about it. Although I think one thing that's interesting is that the discussion around basic income is usually, can we afford it?
[0:12:23] You know, economists talk about, well, this is the GDP and this is the tax rate and this is how much it costs to live on and we can't afford it. The thing that's interesting to me is that if we need it, we're guaranteed to be able to afford it. If the robots take all the jobs, what people forget is that the cost of services and many goods will go down a lot, maybe to zero in some cases. And so the cost to have a good life can get really, really cheap. And so that's not what we're studying.
[0:12:51] I assume, again, that if the robot revolution happens, we'll be able to afford a basic income. The interesting question, though, is no matter how much people complain about work, they kind of like it. They get at least some fulfillment out of it. They get a sense of purpose. It feels good to be needed. You get a community and a sense of belonging from your job. And what I want to study is when we don't get that from jobs, or our current jobs, what do we build society to get that instead?
[0:13:21] What do jobs look like? Even if we don't get paid, what do people do for status and fulfillment and social connection? And I'm willing to believe that whatever the answer to that is looks really different from jobs today, but I'm sure it'll be something that shares those characteristics. So that's what we're trying to study. And I think although it appeals to a lot of people across the U.S. political spectrum, no one has a good answer to the effects it has on society yet.
[0:14:04] Yeah. So I think there are two problems with news, fake news and filter bubbles. The fake news problem I think we're going to make good progress on. I'm optimistic we can do that. I think there's good algorithmic solutions. I think some of the things Facebook has rolled out recently seem very promising to me. And, you know, most people, once they're told from credible sources, they trust that news is fake, we'll move on.
[0:14:49] And I think we're going to solve that problem, though it was a huge problem in the last presidential election cycle. I think it's still not fully understood what a big problem that was. However, the other problem is the problem of filter bubbles and that people want to read news that reinforces what they believe and they don't want to read news that isn't what they want to believe. And the reason this is a problem is because, you know, an internet service that is trying, that develops an algorithm,
[0:15:20] to try to get you to spend as much time and attention there as possible. The way to do that is to show you things that keep you engaged and that you want to read more of. I heard someone that did a recent study looking at a video site, and they were looking at the videos in France that it shows you when you're watching them. And the algorithm tries to maximize how much time you spend on the site. And no matter what you believe, no matter what you start out interested in, if the algorithm
[0:15:49] It runs its course and tries to keep you as engaged as possible. Even if you start out watching videos of Macron, you end up watching videos of Le Pen. And they'll be different depending on what you believe. But that's where the cycle stops. And so there's this algorithm which is optimized for attention. And this is what has decided to get people to stay most engaged. I don't think that helps bring people together because there's very different videos of Le Pen, about Le Pen, depending on what you believe.
[0:16:19] That is the thing that gets people engaged. And so how we break that and what we do about the fact that the thing that people want and the thing that keep people engaged is not necessarily the thing that sort of helps us see the whole picture and bring us together as a society. That's a hard problem. There's people working on it, but I think it will mean changing what we mean when we say algorithms optimizing for up and to the right. Sure.
[0:16:57] Yeah, so Neuralink shares the building with OpenAI. It's a company building a neural interface, and the idea is to put very tiny electrodes into a brain, a human brain, so that you can have a high bandwidth connection between a brain and a computer. The belief here is that developing AI is only part of the problem. For the world to go well for humans, we need humans and AI to merge to some degree, and one way to do this.
[0:17:26] is with a direct connection where the computer is sort of directly connected to the brain and in service of the human brain's desires. And it could be quite wild. I mean, if you could truly connect your brain to the internet, you could think of anything and know the answer. And you wouldn't even be consciously aware that you were trying to search. It would just be in your brain. So it'll take a long time. It'll start slowly. I think the most promising direction is to go in through the blood vessels.
[0:17:56] And put very tiny electrodes near a bunch of the neurons. If the electrodes are tiny enough, they don't trigger the immune response of the brain. But there's a lot of work to do. That's an example of the kind of thing you can do in Silicon Valley that's hard to get taken seriously in this room, apparently. Do I think AI is moving too fast? Well, whatever I think about that, it's going to move much faster. I think almost all of the
[0:18:30] The brightest, most ambitious 18-year-olds I meet now going into college are all studying AI. And so we are going to have a flood of experts in this field. And it's like the third derivative is quite positive. So it's going to happen really fast. And I think the reason we started OpenAI was to try to get ahead of the curve a little bit and be ready and be thinking about the societal impacts and the safety issues.
[0:19:20] I expect that it will start that way for a short period of time. However, I think one thing people forget about technology is how quickly it equalizes. You know, when the very first computers came out, only the richest people could afford them. When the very first iPhone came out, it was still quite expensive but more affordable. At this point, there are $25 Android phones available that are pretty great, and access to the Internet is relatively affordable,
[0:19:51] Even when I go travel and visit places that are desperately poor, the people still have smartphones and data plans and they learn on them and they can access, they have pretty much the same phone I do and pretty much access to the same internet that I do. And so Bill Gates and people way below the poverty line have the same technology and the same access to information. And this is an amazing thing. I think there is a way in which technology
[0:20:19] It creates wealth inequality, and we can talk about that in just a second. But I think there's an important way in which if we do it the right way, technology is an incredible equalizer of opportunity. So I think over time, it probably equalizes more than it hurts, right? Everybody could be equally smart, which would be huge for equality. What I worry about with inequality is the increasing leverage that companies have.
[0:20:50] So we are in the age of these hyperscale companies where you'll have a small number of companies that wield just massive wealth and market power. And that is different than technology. That's more of a monopoly effect where there's a huge network effect and all the value accrues to this one place. So I think that's different than the issues of a particular technology which does get democratized over time. And I think that's the more important thing to think about from an inequality perspective. Sure.
[0:21:36] Well, look, if only one company has the secret for AI, that's really bad. Which is part of the reason, half the reason we started OpenAI, is we want this to be democratized and available to everyone. We're a non-profit, we're going to do this thing and we're going to give away the benefits to everybody. Because I think this is the defining inequality potential of our time. And it's really bad for it to be locked up in one company. So, there's a question though, which is even if we release the code,
[0:22:16] Or even not release the code, but release the product someday. Or give the benefits out to everybody someday. Is there still a problem where only having a lot of data matters or only a lot of computing power matters? I think the data advantage is somewhat overstated. It's in Google, Facebook, anybody else's. It's in their benefit to say you need a huge amount of data because then every good AI person wants to go work there. But there's existence proof in humans that we can learn with small amounts of data. And I expect that when we find the algorithm,
[0:22:48] It'll work with less data than we think. Compute power is a bigger issue. And we do think about how to democratize that. But it certainly, you know, if an oracle told me that the race to AI was going to be won by the United States government because they can spend more money on computers than anybody else, I wouldn't be shocked. So that's an issue there I worry more about. I hope we're not just doing drugs and yoga all day. That's not what I want to be doing.
[0:23:29] You know, I think the answer is people will do different things. There are some people that want to be hippies. There are some people that want to start startups. And I think the easy-to-make mental error when thinking about this world is to assume A, that it'll look a lot like the world looks today, and B, that most people want to do the same thing. Again, human desire, limitless. People's desire to get rich, limitless. People will find new ways to compete and get status and feel superior to each other.
[0:23:58] And some people just do drugs and play video games. And although I think the world will change a lot, I don't expect human nature to change that dramatically. Look, I think we talked about one important thing already, which is what do you do about filter bubbles and fake news? And I think technology needs to address that. I think another thing is people want agency over their own lives and they don't like the feeling that
[0:24:49] The small number of people that control technology control everything. And so I think bringing a wider segment of the world into this conversation and into the decisions we're going to make now is super important. And then I think there's a lot of things that are not about technology but about our government. I think in the U.S. in particular, as politics were driven more and more by money, politicians spent more and more of their time focused on the 1%.
[0:25:19] And none on the 99% because we had all the money and they needed that. And so most people got really ignored by the political system in the last, let's say, 15 years in the US. And I think part of what you heard here was, or saw in our case, was an uprising there. And people want to be heard. People want to be involved in deciding what the future is going to be like. And I think our government has a real responsibility to do that. Tech can help somewhat, but that's more of a society.
[0:25:49] Change that we need. And then finally, I think the wealth inequality issue is just going to be huge. It seems to me crazy that we don't have a large-scale project to eliminate poverty in the US. And I think we could take that on today, and we don't. I think people are angry about that. I think there are also just a lot of people who don't want change, and they don't want
[0:26:21] Who has thought up the group?
[0:27:11] I believe that we will reach one day singularity and that we can
[0:27:54] Do you mean like human machine merge or do you mean machines that are smarter than humans? Yes, for sure. I actually believe both. But I'm not even sure it's that far away. And I think there are already machines that are smarter than humans in a lot of specific areas. I think some of the recent results around perception.
[0:28:26] You know, listening to really difficult audio files, really difficult image processing. I think already computers can do things there that humans cannot. I expect that to keep going. You know, the algorithm for cognition may take a while, but there's a lot of smart people working on it. Unless you believe something sort of non-standard, like if you believe the entire brain is quantum coherent, maybe then...
[0:28:57] True AI is a long way off. But I think it's probably not. And I think we will figure it out at some point. My definition of the singularity is a little bit different, which is sort of the human machine merge. And I think there's an argument that that's already behind us, more than we realize. It's already happened. I think smartphones sort of did it.
[0:29:27] When you walk through a crowd now, most people hold their phone. They get uncomfortable if they can't hold their phone. And these phones tell us what to do. They tell us where to go, when they buzz, we go do something. Algorithms decide what to show us on Facebook or Google or something else. And those algorithms decide if we're going to be happy or sad, what we're going to do, what we're going to think, who we're going to vote for. So there is this very real way in which this sort of cloud super brain is already controlling us.
[0:29:56] And that seems very singularity-like and that it may have gotten out of control in a way we don't yet understand. But looking back in a few more years, it'll be obvious to us that it's happened. In the last 10 plus years, we've been looking at Silicon Valley and European companies versus the U.S. In 10 years' time, where do you think companies will be moving to? Well, I hope Silicon Valley, but I worry we can't get out of our own way with housing and a few other issues.
[0:30:28] You know, I go to China a lot, and it's amazing what's happening there. But what I think the most is that there will just be a lot of places doing really impressive work. And that in 10 or 20 years, there won't be this one place that is responsible for 90% of the market cap created. We're thinking about it this year, or next year. We're thinking about it next year. But we may not. We always ask founders that come from other countries if...
[0:31:07] They were happy to come to Silicon Valley or if they wish we had opened an office in their place. And almost everyone says they were really happy to come to Silicon Valley for three months. So that's where we've dragged our feet so far. And if we keep hearing that and if we keep being able to get the best companies to come, it's only three months. If we keep getting the best companies able to come, we might not. But we are actively considering it for the first time. China would be tempting just because it's so...
[0:31:42] What do you consider a startup studio? What do you consider in the U.S.?
[0:32:21] There's a long history of it not working there. I think it's something different about the culture that I don't understand. It's a great observation. It's definitely happened. The giant tech companies are sort of sovereign states at this point. And it's easy to imagine a world in which they're more powerful than sovereign states.
[0:33:02] Even what Combinator is surprisingly powerful, much more so than I expected a few years ago. I think that trend will continue, and I think it's time for a conversation about how all of us as a society feel about that and what, if anything, should be done. You have to go back pretty far in history to see companies that have been this powerful. Should we stop and eat and hang out? Oh, one more. All right.
[0:33:37] I mean, it's really hard. I wonder about that question a lot. And you have two very strong opposing forces. So there's been this trend towards globalization, I think led by technology companies.
[0:34:16] Feeling that the world is getting smaller and that the internet makes us all live in this one, I don't even know what you want to call it, like, made-us society where we're all connected. On the other hand, you've seen in the last few years this real populism streak, and it feels like the tide is kind of going out on globalization. And, you know, people say the EU is falling apart, people say the US is going to cut off all of its partnerships with other countries.
[0:34:44] So there's one big thrust vector in each direction pushing against each other. There's the internet bringing everyone together and there's this effect of no one can really ever say precisely if it's globalization or technology. They feel the same to people. The job loss of globalization and technology feels the same to people. I actually think it's completely orthogonal. But whatever this vague enemy is that's destroying society, destroying culture, destroying jobs that people are upset about.
[0:35:14] There's a feeling that the way to resist that is to put up bigger borders. And so you have these two opposing forces and I don't know which one's going to win. It feels like it's still up in the air. I think in the long term we move more towards sort of a unified worldwide society. But I'm not sure about that. And certainly in the medium term it feels like a coin flip. All right, one more. How do I feel about taxing robots? I think that...
[0:35:49] Sounds good in theory, but when you really think through it, there's a long list of real problems. What I think should be taxed is wealth. Generated by robots, humans, whatever. I think that's what a good tax system would be like. I have a bunch of objections to the current US tax system, but I think we will, like it or not, move more towards a system where we tax wealth, however it's generated. Monopoly company, robot, labor, whatever.
[0:36:19] And that will be more fair. In the US, for a long time, we didn't even have an income tax. It was just a land tax, which is a version of a wealth tax, really. And that is personally what I believe is the closest thing to the optimal system. All right. Thank you all very much. Thanks for having me. Absolutely.